Mert Kosan successfully passed his PhD proposal on the topic “Transparent Representation Learning for Attributed Graphs”.

Committee: Ambuj Singh (Chair), Francesco Bullo, Xifeng Yan


Graph data show relationships between entities in a variety of domains including but not limited to communication, social, and interaction networks. Representation learning makes graph data easier to use for graph tasks such as graph classification, link prediction, and clustering. The decisions on graphs depend on complex patterns combining rich structural and attribute data. Explaining these decisions made by representation learning models for high-stakes applications (e.g., drug discovery and fraud detection) is critical for increasing transparency and guiding improvements. During my proposal, I will motivate why graphs and representation learning are useful for machine learning and why we should work on transparent models. Then, I will review my research about explainable machine learning models on different tasks and how to achieve fairness in clustering. Lastly, I will talk about how machine learning models can analyze human behaviors such as decision-making under uncertainty.